Appendix A — How this edition was made

This edition was rebuilt from the scans, line by line, in October 2026. Every table was then checked by eye against the scanned page. This appendix records how that was done and what it found, so that anyone relying on a 1988 figure knows how far to trust it.

Sources

The source is a scan of the 1990 typescript report, made in January 2013 on a Fujitsu ScanSnap at 400 dots per inch, in two PDF files: 1988 survey of ringing Pt 1 (57 pages, report pages 1–56 and the contents) and Pt 2 (48 pages, report pages 57–104). The scans are not in the repository, because each is larger than the repository’s 1 MB limit on any one file.

An earlier transcription, made with Google Gemini from the same scans, served as the first version of this edition. It proved unreliable as a record. Gemini had rewritten as well as transcribed: it left out about 6,000 words of the original, mostly whole sentences and asides; it reformatted and retitled tables; and it changed figures. In Table 18.1, for example, it moved values between columns, and in Table 10.2 it gave 12 per cent of employed ringers as professionals in law, accountancy, statistics and computing, where the report says 6.

Method

  1. Each page of both scans was read with Tesseract, an open-source optical character recognition program, so that every word came from the page image and not from a language model.
  2. The text was compared word by word with the Gemini version. Where they differed, the page image settled it. Scanning errors were corrected (for example “belis” to “bells”, “S-bell” to “5-bell”).
  3. Every table was retyped from Tesseract’s output into a regular grid and then checked cell by cell against the page image. Where a printed table sets several distributions side by side, they are stacked here so that each row is one category; an editor’s note says so under the table.
  4. Arithmetic printed in the report was re-run where it could be. For example, the net change column of Table 9.5 reproduces exactly from the other columns, in every row.

The result is 58 tables (43 numbered in the original, the rest set in the text) and 1588 table cells, held in data/reference/survey_1988/ and described in Appendix B.

What the checks found in the 1988 report

These are slips in the original, not in the transcription. Each has an editor’s note at the point where it occurs. None has been corrected in the text or the data, because a 2026 reader needs to see what 1988 readers saw.

Where What is printed Why it matters
Chapters 2, 4, 9 The number of towers in the population is given as 5425, 5245 and 5250 The estimate of 41,000 ringers multiplies by 5245; with 5425 the same method gives about 43,000
Chapter 3 62.8 per cent of bells rung regularly; the counts printed give 65.2 per cent The headline figure may reflect returns added later (Chapter 1 says figures were updated)
Chapter 4 62.5 per cent, against 62.8 in Chapter 3 Minor inconsistency
Table 9.4 against Table 5.4 Five figures for 10- and 12-bell towers differ, by up to nine points Same question, same respondents
Table 10.8 against Table 7.5 Tower captains’ view of Branch social events: 47 and 3 per cent, against 43 and 7 Same question, same respondents
Chapter 7 against Table 17.1 60 and 63 tower captains in the youngest and oldest groups, against 67 and 69 Different bases, probably before and after late returns
Table 13.2 145 per cent of contests had 1–4 entries Probably 14; with 14 the column sums to 100
Table 19.1 One row sums to 95 Rounding or a misprint
Chapter 9 Large cities named as both worst and best for net change The table shows suburbs and inner cities worst
Chapters 9, 11, 14 Some cross-references point to the wrong chapter Corrected references are given in the notes

What remains uncertain

Pages 1–37 were checked against the scans in two passes, one automatic and one by eye for every table and every page with figures. Prose on the other pages of that range was checked against Tesseract’s text and Gemini’s only. A misread word may survive there, though no figure should. Pages 38–104 were all read against the page image.

Changes from the printed layout

  • Underlining, the typewriter’s way of marking a title, is shown as italics.
  • Rows of asterisks that separate an article’s introduction from its body are shown as a rule.
  • The two maps in Chapter 2 are the scanned images, cropped.
  • The answers to the quiz in Chapter 22 are hidden until you click to show them.